The reliability and measurement errors associated with the assessment of plaque, calculus, gingival bleeding and loss of attachment in two cohorts of subjects were investigated. Each type of measurement was replicated by the same examiner in two cohorts (29 Norwegians under office conditions, and 35 Sri Lankans under field conditions). All replicates were performed within a 24-hour recall period. The results showed that these examiners were able to perform more reliably in an office setting than under field conditions. The examiners were best able to score gingival bleeding and loss of attachment, and were less consistent when scoring calculus and plaque. The estimated effect of measurement error for these indices in a longitudinal study would be rather small provided whole-mouth average scores are used to assess changes in disease levels because examiner differences explained a relatively small percentage of the variation in these scores.